User interface

[drdhaval2785]

@rakeshvar

Wouldn't it be a good idea to have a frontend for your code - so that people with not too much technical knowledge (like me) will be able to test?

maybe appspot for trial?

[rakeshvar]

Dr Dhaval,
This currently works only for non-contiguous scripts like South Indian ones, and Gujrati, Oriya etc. This framework does not work for languages with shirorekha!
I am busy with the Machine Learning part (which is almost ready), I need a volunteer to build the front-end, as I am not an expert on such things, I also need a website, and back-end infrastructure.

• Rakeshvar

[drdhaval2785]

@rakeshvar
Any progress in this regards ?
At least a bunch of installation files would be a welcome step.
Right now I am not able to test the OCR.

If you can make a video of how-to use this machine - it would be of great help.

I read an article regarding issue of shirorekha in OCR and how it can be circumvented.
The concept crux was something like this.

1. Let's presume the height of shirorekha as x.
2. We take the values of black dots on y - axis above a given point and plot it on X-axis.
3. If the total is >x, we retain it.
4. If the total is <=x, we remove it (thereby removing the shirorekha and getting separate letters).

Let me explain it with an example.

In this -
cross section 'a' - has value 3x (3 black intersection).
cross section 'b' - has value x (only shirorekha intersection).
cross section 'c' - has value 4x (4 intersections).

So we remove the black points where the value is only 'x' or maybe less than that.
After that we will get separate letters - without shirorekha.

If I get access to the full paper - I will share.
Hope it was useful

[drdhaval2785]

http://research.ijcaonline.org/volume39/number6/pxc3877076.pdf

This is the paper I was referring to.

[rakeshvar]

Please check your gmail.